Two passes, and why it matters
GIF allows at most 256 colours in a frame. Video routinely contains hundreds of thousands. Something has to decide which 256 survive, and that decision is the entire difference between a GIF that looks like the source and one that looks like a fax of it.
The naive approach uses a fixed, generic palette — an even spread across the colour cube that matches nothing in particular. Skin tones go blotchy, gradients band into visible stripes, and anything with a dominant hue turns muddy.
This tool instead runs palettegen across the frames you selected, producing a palette optimised for the colours actually present, then encodes with paletteuse against it. If your clip is mostly a blue sky and a green field, nearly all 256 slots go to blues and greens. The cost is a second pass over the footage; the benefit is obvious the moment you compare the two.
Why GIF files are so large
Modern video compression is built on the observation that consecutive frames are nearly identical. Store one complete frame, then describe the next twenty as small differences from it, and you get the enormous compression ratios that make streaming possible.
GIF does none of this. Each frame is compressed on its own with LZW, a general-purpose algorithm designed in 1984 for images with large flat areas. It has no motion model at all. The practical result is that a clip which is 2MB as MP4 can be 8MB or more as GIF — the same pictures, four times the bytes.
Duration, frame rate and width all multiply into the total, which is why the estimate updates as you change them and why the duration is capped.
Choosing settings
- 8 fps — slow or largely static footage; smallest files
- 12 fps — the usual default; smooth enough for most motion
- 15 fps — noticeably smoother, roughly 25% larger
- 20–25 fps — only worth it for genuinely fast motion
Width behaves the same way but more steeply, because the pixel count grows with the square. Going from 320px to 640px roughly quadruples the data per frame. Pick the smallest width at which the content is still readable.
When not to use a GIF
If the destination accepts video, use video. An MP4 of the same clip will be smaller, sharper, support audio, and play everywhere a GIF does. GIF earns its place in the specific cases where a still image is expected but motion is wanted — inline in documentation, in a README, in an email, or on a platform that treats video differently from images.